Search

This paper proposes an automated active fire detection framework using Sentinel-2 imagery. The framework is made up of three basic parts including data collection and preprocessing, deep-learning-based active fire detection, and final product generation modules. The active fire detection module is developed on a specifically designed dual-domain channel-position attention (DCPA)+HRNetV2 model and a dataset with semi-manually annotated active fire samples is constructed over wildfires that commenced on the east coast of Australia and the west coast of the United States in 2019–2020 for the training process. This dataset can be used as a benchmark for other deep-learning-based algorithms to improve active fire detection accuracy. The performance of active fire detection is evaluated regarding the detection accuracy of deep-learning-based models and the processing efficiency of the whole framework. Results indicate that the DCPA and HRNetV2 combination surpasses DeepLabV3 and HRNetV2 models for active fire detection. In addition, the automated framework can deliver active fire detection results of Sentinel-2 inputs with coverage of about 12,000 km2 (including data download) in less than 6 min, where average intersections over union (IoUs) of 70.4% and 71.9% were achieved in tests over Australia and the United States, respectively. Concepts in this framework can be further applied to other remote sensing sensors with data acquisitions in SWIR-NIR-Red ranges and can serve as a powerful tool to deal with large volumes of high-resolution data used in future fire monitoring systems and as a cost-efficient resource in support of governments and fire service agencies that need timely, optimized firefighting plans. [ABSTRACT FROM AUTHOR]

In this paper, a deep learning long-short-term memory (LSTM) method is applied to the forecasting of the critical frequency of the ionosphere F2 layer (foF2). Hourly values of foF2 from 10 ionospheric stations in China and Australia (based on availability) from 2006 to 2019 are used for training and verifying. While 2015 and 2019 are exclusive for verifying the forecasting accuracy. The inputs of the LSTM model are sequential data for the previous values, which include local time (LT), day number, solar zenith angle, the sunspot number (SSN), the daily F10.7 solar flux, geomagnetic the Ap and Kp indices, geographic coordinates, neutral winds, and the observed value of foF2 at the previous moment. To evaluate the forecasting ability of the deep learning LSTM model, two different neural network forecasting models: a back-propagation neural network (BPNN) and a genetic algorithm optimized backpropagation neural network (GABP) were established for comparative analysis. The foF2 parameters were forecasted under geomagnetic quiet and geomagnetic disturbed conditions during solar activity maximum (2015) and minimum (2019), respectively. The forecasting results of these models are compared with those of the international reference ionosphere model (IRI2016) and the measurements. The diurnal and seasonal variations of foF2 for the 4 models were compared and analyzed from 8 selected verification stations. The forecasting results reveal that the deep learning LSTM model presents the optimal performance of all models in forecasting the time series of foF2, while the IRI2016 model has the poorest forecasting performance, and the BPNN model and GABP model are between two of them. [ABSTRACT FROM AUTHOR]

The prediction of sea level rise is extremely important for improved future climate change mitigation and adaptation strategies. This study uses a hybrid convolutional neural Network (CNN) and a bidirectional long short-term (BiLSTM) model with successive variational mode decomposition (SVMD) to predict the absolute sea level for two study sites in Australia (Port Kembla and Milner Bay). More importantly, the sea level measurements using a tide gauge were corrected using Global Navigation Satellite System (GNSS) measurements of the vertical land movement (VLM). The SVMD-CNN-BiLSTM model was benchmarked by a multi-layer perceptron (MLP), support vector regression (SVR) and gradient boosting (GB). The SVMD-CNN-BiLSTM model outperformed all the comparative models with high correlation values of more than 0.95 for Port Kembla and Milner Bay. Similarly, the SVMD-CNN-BiLSTM model achieved the highest values for the Willmott index, the Nash–Sutcliffe index and the Legates and McCabe index for both study sites. The projected linear trend showed the expected annual mean sea rise for 2030. Using the current trend, Port Kembla was projected to have an MSL value of 1.03 m with a rate rise of approx. 4.5 mm/year. The rate of the MSL for Milner Bay was comparatively lower with a value of approx. 2.75 mm/year and an expected MSL value of 1.27 m for the year 2030. [ABSTRACT FROM AUTHOR]

Satellite imagery is the only feasible approach to annual monitoring and reporting on land cover change. Unfortunately, conventional pixel-based classification methods based on spectral response only (e.g., using random forests algorithms) have shown a lack of spatial and temporal stability due, for instance, to variability between individual pixels and changes in vegetation condition, respectively. Machine learning methods that consider spatial patterns in addition to reflectance can address some of these issues. In this study, a convolutional neural network (CNN) model, U-Net, was trained for a 500 km × 500 km region in southeast Australia using annual Landsat geomedian data for the relatively dry and wet years of 2018 and 2020, respectively. The label data for model training was an eight-class classification inferred from a static land-use map, enhanced using forest-extent mapping. Here, we wished to analyse the benefits of CNN-based land cover mapping and reporting over 34 years (1987–2020). We used the trained model to generate annual land cover maps for a 100 km × 100 km tile near the Australian Capital Territory. We developed innovative diagnostic methods to assess spatial and temporal stability, analysed how the CNN method differs from pixel-based mapping and compared it with two reference land cover products available for some years. Our U-Net CNN results showed better spatial and temporal stability with, respectively, overall accuracy of 89% verses 82% for reference pixel-based mapping, and 76% of pixels unchanged over 33 years. This gave a clearer insight into where and when land cover change occurred compared to reference mapping, where only 30% of pixels were conserved. Remaining issues include edge effects associated with the CNN method and a limited ability to distinguish some land cover types (e.g., broadacre crops vs. pasture). We conclude that the CNN model was better for understanding broad-scale land cover change, use in environmental accounting and natural resource management, whereas pixel-based approaches sometimes more accurately represented small-scale changes in land cover. [ABSTRACT FROM AUTHOR]

Accurate measurements of the structural characteristics of trees such as height, diameter, sweep and taper are an important part of forest inventories in managed forests and commercial plantations. Both terrestrial and aerial LiDAR are currently employed to produce pointcloud data from which inventory metrics can be determined. Terrestrial/ground-based scanning typically provides pointclouds resolutions of many thousands of points per m 2 from which tree stems can be observed and inventory measurements made directly, whereas typical resolutions from aerial scanning (tens of points per m 2 ) require inventory metrics to be regressed from LiDAR variables using inventory reference data collected from the ground. Recent developments in miniaturised LiDAR sensors are enabling aerial capture of pointclouds from low-flying aircraft at high-resolutions (hundreds of points per m 2 ) from which tree stem information starts to become directly visible, enabling the possibility for plot-scale inventories that do not require access to the ground. In this paper, we develop new approaches to automated tree detection, segmentation and stem reconstruction using algorithms based on deep supervised machine learning which are designed for use with aerially acquired high-resolution LiDAR pointclouds. Our approach is able to isolate individual trees, determine tree stem points and further build a segmented model of the main tree stem that encompasses tree height, diameter, taper, and sweep. Through the use of deep learning models, our approach is able to adapt to variations in pointcloud densities and partial occlusions that are particularly prevalent when data is captured from the air. We present results of our algorithms using high-resolution LiDAR pointclouds captured from a helicopter over two Radiata pine forests in NSW, Australia. [ABSTRACT FROM AUTHOR]

Land cover mapping from satellite images has progressed from visual and statistical approaches to Random Forests (RFs) and, more recently, advanced image recognition techniques such as convolutional neural networks (CNNs). CNNs have a conceptual benefit over RFs in recognising spatial feature context, but potentially at the cost of reduced spatial detail. We tested the use of CNNs for improved land cover mapping based on Landsat data, compared with RFs, for a study area of approximately 500 km × 500 km in southeastern Australia. Landsat 8 geomedian composite surface reflectances were available for 2018. Label data were a simple nine-member land cover classification derived from reference land use mapping (Catchment Scale Land Use of Australia—CLUM), and further enhanced by using custom forest extent mapping (Forests of Australia). Experiments were undertaken testing U-Net CNN for segmentation of Landsat 8 geomedian imagery to determine the optimal combination of input Landsat 8 bands. The results were compared with those from a simple autoencoder as well as an RF model. Segmentation test results for the best performing U-Net CNN models produced an overall accuracy of 79% and weighted-mean F1 score of 77% (9 band input) or 76% (6 band input) for a simple nine-member land cover classification, compared with 73% and 68% (6 band input), respectively, for the best RF model. We conclude that U-Net CNN models can generate annual land cover maps with good accuracy from proxy training data, and can also be used for quality control or improvement of existing land cover products. [ABSTRACT FROM AUTHOR]

Large damages and losses resulting from floods are widely reported across the globe. Thus, the identification of the flood-prone zones on a flood susceptibility map is very essential. To do so, 13 conditioning factors influencing the flood occurrence in Brisbane river catchment in Australia (i.e., topographic, water-related, geological, and land use factors) were acquired for further processing and modeling. In this study, artificial neural networks (ANN), deep learning neural networks (DLNN), and optimized DLNN using particle swarm optimization (PSO) were exploited to predict and estimate the susceptible areas to the future floods. The significance of the conditioning factors analysis for the region highlighted that altitude, distance from river, sediment transport index (STI), and slope played the most important roles, whereas stream power index (SPI) did not contribute to the hazardous situation. The performance of the models was evaluated against the statistical tests such as sensitivity, specificity, the area under curve (AUC), and true skill statistic (TSS). DLNN and PSO-DLNN models obtained the highest values of sensitivity (0.99) for the training stage to compare with ANN. Moreover, the validations of specificity and TSS for PSO-DLNN recorded the highest values of 0.98 and 0.90, respectively, compared with those obtained by ANN and DLNN. The best accuracies by AUC were evaluated in PSO-DLNN (0.99 in training and 0.98 in testing datasets), followed by DLNN and ANN. Therefore, the optimized PSO-DLNN proved its robustness to compare with other methods. [ABSTRACT FROM AUTHOR]

Digital agriculture services can greatly assist growers to monitor their fields and optimize their use throughout the growing season. Thus, knowing the exact location of fields and their boundaries is a prerequisite. Unlike property boundaries, which are recorded in local council or title records, field boundaries are not historically recorded. As a result, digital services currently ask their users to manually draw their field, which is time-consuming and creates disincentives. Here, we present a generalized method, hereafter referred to as DECODE (DEtect, COnsolidate, and DElinetate), that automatically extracts accurate field boundary data from satellite imagery using deep learning based on spatial, spectral, and temporal cues. We introduce a new convolutional neural network (FracTAL ResUNet) as well as two uncertainty metrics to characterize the confidence of the field detection and field delineation processes. We finally propose a new methodology to compare and summarize field-based accuracy metrics. To demonstrate the performance and scalability of our method, we extracted fields across the Australian grains zone with a pixel-based accuracy of 0.87 and a field-based accuracy of up to 0.88 depending on the metric. We also trained a model on data from South Africa instead of Australia and found it transferred well to unseen Australian landscapes. We conclude that the accuracy, scalability and transferability of DECODE shows that large-scale field boundary extraction based on deep learning has reached operational maturity. This opens the door to new agricultural services that provide routine, near-real time field-based analytics. [ABSTRACT FROM AUTHOR]

Precision viticulture benefits from the accurate detection of vineyard vegetation from remote sensing, without a priori knowledge of vine locations. Vineyard detection enables efficient, and potentially automated, derivation of spatial measures such as length and area of crop, and hence required volumes of water, fertilizer, and other resources. Machine learning techniques have provided significant advancements in recent years in the areas of image segmentation, classification, and object detection, with neural networks shown to perform well in the detection of vineyards and other crops. However, what has not been extensively quantitatively examined is the extent to which the initial choice of input imagery impacts detection/segmentation accuracy. Here, we use a standard deep convolutional neural network (CNN) to detect and segment vineyards across Australia using DigitalGlobe Worldview-2 images at ∼50 cm (panchromatic) and ∼2 m (multispectral) spatial resolution. A quantitative assessment of the variation in model performance with input parameters during model training is presented from a remote sensing perspective, with combinations of panchromatic, multispectral, pan-sharpened multispectral, and the spectral Normalised Difference Vegetation Index (NDVI) considered. The impact of image acquisition parameters—namely, the off-nadir angle and solar elevation angle—on the quality of pan-sharpening is also assessed. The results are synthesised into a 'recipe' for optimising the accuracy of vineyard segmentation, which can provide a guide to others aiming to implement or improve automated crop detection and classification. [ABSTRACT FROM AUTHOR]